Combine has spent nearly a decade helping organizations navigate technology change — implementations, integrations, transformations, the works. We've seen what separates technology adoption that sticks from technology adoption that fades. And when AI started to matter seriously — not as a toy, but as a capability that could reshape how organizations operate — we paid close attention.
What we saw was a familiar pattern playing out at unusual speed. Organizations were deploying AI tools faster than they were building the organizational capability to use them well. The tools were good. The adoption infrastructure was missing.
The gap we kept seeing
Every organization we worked with in 2024 was somewhere on the same spectrum. On one end: "We have a ChatGPT license and nobody really uses it." On the other: "We have a few power users who are doing incredible things, but we can't figure out how to spread that across the org." Almost nobody was in the middle — systematically deploying AI in a way that was consistent, measurable, and improving over time.
The tools on the market weren't solving this. They were either generic AI assistants (powerful but unadapted) or narrow workflow tools (useful for one thing, unusable for another). Nobody was building the layer between the model and the organization — the configuration, the coaching, the measurement, the iteration.
What we decided to build
We started building Luz in early 2025. The brief was simple: build the platform we wish existed when we were advising organizations on AI adoption. It needed to be configurable for each org, not generic. It needed to coach in real time, not in workshops. It needed to measure what mattered, not just usage counts. And it needed to get better over time, not just sit there.
- A configuration layer that encodes org structure, policies, and context into every AI interaction
- A real-time coaching system that improves prompt quality at the moment it matters
- An analytics layer that measures adoption quality, not just volume
- A direction module that turns data into strategic recommendations
- A workflow system that makes institutional knowledge accessible to everyone
“We didn't build Luz because AI is exciting. We built it because the organizational work of deploying AI well was being skipped — and we knew how to fix that.”
Where we are now
Luz is now in early access with a select group of organizations. We're onboarding deliberately — we want to work closely with each team, understand their specific context, and make sure the platform is actually working before we scale. If you're building an organization where AI matters, we'd like to talk.
We're a small team with a clear thesis and a lot of conviction. If the problem we've described here resonates — if you've lived that gap between AI potential and AI reality in your own organization — we think you'll find Luz worth your time.